Download full text - pdf - reports on mathematical logic

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Structure (mathematical logic) wikipedia , lookup

Fundamental theorem of algebra wikipedia , lookup

Propositional calculus wikipedia , lookup

Transcript
REPORTS ON MATHEMATICAL LOGIC
51 (2016), 105–131
doi:10.4467/20842589RM.16.008.5285
Gemma ROBLES
THE QUASI-RELEVANT
3-VALUED LOGIC RM3
AND SOME OF ITS SUBLOGICS
LACKING THE VARIABLE-SHARING
PROPERTY
A b s t r a c t. The logic RM3 is the 3-valued extension of the logic
R-Mingle (RM). RM (and so, RM3) does not have the variablesharing property (vsp), but RM3 (and so, RM) lacks the more
“offending” “paradoxes of relevance”, such as A → (B → A) or
¬A → (A → B). Thus, RM and RM3 can be useful when “some
relevance”, but not the full vsp, is needed. Sublogics of RM3 with
the vsp are well known, but this is not the case with those lacking
this property. The first aim of this paper is to define an ample
family of sublogics of RM3 without the vsp. The second one is
to provide these sublogics and RM3 itself with a general RoutleyMeyer semantics, that is, the semantics devised for relevant logics
in the early seventies of the past century.
Received 9 December 2015
Keywords and phrases: RM3; R-Mingle; relevant logics; quasi-relevant logics; RoutleyMeyer semantics; relational semantics; substructural logics.
106
GEMMA ROBLES
1. Introduction
The first aim of this paper is to define an ample family of sublogics of RM3
without the variable-sharing property. The second one is to endow these
sublogics and RM3 itself with a general Routley-Meyer semantics.
As it is well known (cf. [1]), a necessary property of any relevant logic
S is the following:
Definition 1.1 (Variable-sharing property – vsp). If A → B is a theorem of S, then A and B share at least a propositional variable.
In [1] (§22.1.3), it is shown that the logic of relevance R, and so any
logic included in it (as, for example, the logic of entailment E or the logic
of ticket entailment T), has the vsp. There is, however, an interesting logic
related to R, the logic R-Mingle (RM), that does not have the vsp. The
logic RM is axiomatized by adding the axiom mingle A → (A → A) to the
logic of relevance R (the axiom mingle is labelled A9 below; cf. Definition
2.6). In RM “paradoxes of relevance” (i.e., conditionals in which antecedent
and consequent do not share propositional variables) are almost immediate
(cf. [1], §29.5). But, despite this fact, RM has been given considerable
attention in volume 1 of Entailment (cf. [1], §29.3) because it lacks the
most “offensive” paradoxes of relevance (A → (B → A) or ¬A → (A → B),
for example) and, in fact, it has the following property, akin to the vsp (cf.
[1], p. 417):
Definition 1.2 (Quasi-relevance property). If A → B is a theorem,
then either (1) A and B share at least a propositional variable or (2) both
¬A and B are theorems.
(Cf. Definition 2.1. below on the logical languages used in the paper.)
Thus, as Meyer puts it (cf. [1], p.393): “Sometimes one doesn’t need the
whole relevance principle and, in those occasions, RM is good enough, when
some relevance is desirable.” (By the ‘whole relevance principle’, Meyer
refers to the vsp.)
The logic RM3 is the 3-valued extension of RM and its has some of the
more important properties of the latter logic such as the quasi-relevance
property (cf. Section 4 of the appendix). The logic RM3 can be axiomatized
by adding any of the axioms A ∨ (A → B) or ¬A → [A ∨ (A → B)]
to RM. (These axioms are labelled t18 and A11, respectively, below. Cf.
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
107
Proposition 4.1.) So, RM3 can be axiomatized by adding A9 A → (A → A)
and either t18 A∨(A → B) or A11 ¬A → [A∨(A → B)] to R (cf. Section 3
of the appendix). Concerning these axioms, it should be noted that t18 can
be added to R, the vsp being preserved, as shown in the 5-valued relevant
logic CL, one axiom of which is t18. The logic CL is an extension of R
axiomatizing Meyer’s Crystal lattice CL, and it has the vsp (cf. [5], p. 114,
ff.). The axiom mingle A9, however, causes the breaking of the vsp when
added to weak systems with this property (cf. [12]). RM3 has been given
an algebraic semantics in [7] (cf. also [1], §29.4). Also, it has been endowed
in [2] with a Belnap-Dunn “bivalent” type semantics as well as a 2 set-up
model structure of the kind defined in Routley-Meyer semantics (cf. [14]
and references therein).
Next, we describe the aims of the paper. Sublogics of RM and RM3
with the vsp are well known: they include the logic of relevance R and their
sublogics such as the logic of entailment E, the logic of ticket entailment
T, not to mention weak relevant logics among which Brady’s logic DR
and its sublogics are undoubtedly the more important (cf. [3], [4], [6]).
Nevertheless, sublogics of RM3 without the vsp have not been studied yet,
as far as we know. But these logics are interesting and useful, since, as
remarked above, there are situations where we need “some relevance”, but
not necessarily the full vsp. Consequently, the first aim of this paper is to
define an ample family of sublogics of RM3 without the vsp. The minimal
logic among these is the result of adding to Routley and Meyer’s basic
logic B (cf. [14]; cf. Definition 2.5 below) the characteristic axioms of
RM3, A9 and A11, referred to above, together with the auxiliary axiom
A10 (A ∧ ¬A) → (B ∨ ¬B) labeled “safety” in its rule form in [8], p.
14. The second aim of this paper is to endow these sublogics of RM3
and RM3 itself with a general Routley-Meyer semantics (RM-semantics),
the semantics devised for relevant logics in the early seventies of the past
century (cf. [14]). In this sense, we remark that although there is more
than one semantical approach to RM3, as it was pointed above, we do not
have a general Routley-Meyer semantics for RM3 or any of its sublogics
without the vsp yet, as far as we know.
The paper is organized as follows. In Section 2, the logic BRM3 is
defined, a Routley-Meyer semantics for BRM3 is provided and soundness
is proved w.r.t. this semantics. The label BRM3 stands for ‘Basic nonrelevant logic included in RM3’. In Section 3, it is proved that BRM3 is
108
GEMMA ROBLES
complete w.r.t. the semantics defined in the previous section, by using a
canonical model construction. In Section 4, we define an ample family of
sublogics of BMR3 without the vsp and endow each one of them with a general RM-semantics. The section is ended by providing two alternative ways
of defining models for RM3 in the RM-semantics. In Section 5, we draw
some conclusions from the results obtained and suggest some directions for
further work on the same topic. Finally, we have included an appendix
recording some matrices upon which some proofs in the preceding sections
are based and some other additional material.
2. The logic BRM3 and its semantics
Definition 2.1 (Languages, logics). The propositional language consists of a denumerable set of propositional variables p0 , p1 , ..., pn , ... and
some or all of the following connectives → (conditional), ∧ (conjunction),
∨ (disjunction), and ¬ (negation). The biconditional (↔) and the set of
wffs are defined in the customary way. A, B, C, etc., are metalinguistic variables. From the proof-theoretical point of view, we shall consider propositional logics formulated in the Hilbert-style way, that is, logics axiomatized
by means of a set of axioms (actually, axiom schemes) and a set of rules
of derivation. The notions of ‘proof’ and ‘theorem’ are understood as it is
customary in Hilbert-style axiomatic systems. By `S A, it is indicated that
A is a theorem of S.
Definition 2.2 (Logical matrix). A (logical) matrix is a structure
(V, D, F) where (1) V is a (ordered) set of (truth) values; (2) D is a nonempty proper set of V (the set of designated values); and (3) F is the set of
n-ary functions on V such that for each n-ary connective c (of the propositional language in question), there is a function fc ∈ F : V n → V.
Definition 2.3 (M-interpretations, M-validity). Let M be a matrix
for (a propositional language) L. An M-interpretation I is a function from
the set of all wffs to V according to the functions in F. Then, M A (A is
M-valid; A is valid in the matrix M) iff I(A) ∈ D for all M-intepretations I.
Definition 2.4 (The matrix MRM3). The propositional language consists of the connectives →, ∧, ∨ and ¬. The matrix MRM3 is the structure
109
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
(V, D, F) where (1) V is {0, 1, 2} and it is ordered as shown in the following
diagram:
0
1
2
(2) D = {1, 2}; (3) F = {f→ , f∧ , f∨ , f¬ } and these functions are defined
according to the following tables.
→
0
1
2
0
2
0
0
1
2
1
0
2
2
2
2
¬
2
1
0
∧
0
1
2
0
0
0
0
1
0
1
1
2
0
1
2
∨
0
1
2
0
0
1
2
1
1
1
2
2
2
2
2
The notions of an RM3-interpretation and RM3-validity are defined
according to the general Definition 2.3.
The logic RM3 axiomatizing the matrix MRM3 is recorded in section
3 of the appendix. (A logic S axiomatizes a matrix MS iff for every wff A,
`S A iff MS A.) The logic BRM3 (the label intends to abbreviate ‘Basic
non-relevant logic included in RM3’) is defined from Routley and Meyer’s
basic logic B as follows.
Definition 2.5 (The logics B+ and B). Routley and Meyer’s basic positive logic B+ is formulated with the following axioms and rules of inference
(cf. [13] or [14]).
Axioms
A1. A → A
A2. (A ∧ B) → A / (A ∧ B) → B
A3. [(A → B) ∧ (A → C)] → [A → (B ∧ C)]
A4. A → (A ∨ B) / B → (A ∨ B)
A5. [(A → C) ∧ (B → C)] → [(A ∨ B) → C]
A6. [A ∧ (B ∨ C)] → [(A ∧ B) ∨ (A ∧ C)]
Rules of derivation
Modus Ponens (MP): A & A → B ⇒ B
Adjunction (Adj): A & B ⇒ A ∧ B
Suffixing (Suf): A → B ⇒ (B → C) → (A → C)
Prefixing (Pref): B → C ⇒ (A → B) → (A → C)
110
GEMMA ROBLES
Next, Routley and Meyer’s basic logic B is axiomatized when adding
the following axioms and rule to B+ (cf. [14]):
A7. A → ¬¬A
A8. ¬¬A → A
Contraposition (Con) A → B ⇒ ¬B → ¬A
Definition 2.6 (The logic BRM3). The logic BRM3 is formulated by
adding the following axioms to B:
A9. A → (A → A)
A10. (A ∧ ¬A) → (B ∨ ¬B)
A11. ¬A → [A ∨ (A → B)]
We record some theorems of BRM3 and a couple of facts about this
logic.
Proposition 2.7 (Some theorems of BRM3). The following are theorems and rule of BRM3 (a proof is sketched to the right of each one of
them)
T1. ¬A → B ⇒ ¬B → A
Con, A8
T2. ¬(A ∨ B) ↔ (¬A ∧ ¬B)
A3, A4, Con; A3, A5, Con, A7
T3. ¬(A ∧ B) ↔ (¬A ∨ ¬B)
A3, A4, T1; A3, A4, A8, Con
T4. A → [B → (A ∨ B)]
T5. A → [¬A ∨ (¬A → B)]
A4, A9
A7, A11
Proposition 2.8 (On the axiomatization of BRM3). The logic BRM3
is well axiomatized w.r.t. B+ . That is, given the logic B+ , A7, A8, A9,
A10, A11 and Con are independent of each other.
Proof. See the appendix.
Remark 1 (On BRM3 and vsp). Axiom A10 (A ∧ ¬A) → (B ∨ ¬B)
clearly breaks the variable-sharing property (vsp). So, any extension of
BRM3 lacks the vsp whence all of them (BRM3 included) are not relevant
logics. (A logic is relevant in the minimal sense of the term if it has the
vsp.)
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
111
Next, models and the notion of validity are defined. Then, the soundness theorem is proved.
Definition 2.9 (BRM3-models). A BRM3-model is a structure (K, O,
R, ∗, ) where K is a set, O ⊆ K, R is a ternary relation on K and ∗ is a
unary operation on K subject to the following definitions and postulates
for all a, b, c ∈ K:
d1. a ≤ b =df (∃x ∈ O)Rxab
P1. a ≤ a
P2. (a ≤ b & Rbcd) ⇒ Racd
P3. a ≤ a∗∗
P4. a∗∗ ≤ a
P5. a ≤ b ⇒ b∗ ≤ a∗
P6. Rabc ⇒ (a ≤ c or b ≤ c)
P7. a ≤ a∗ or a∗ ≤ a
P8. Rabc ⇒ (b ≤ a or b ≤ a∗)
Finally, is a relation from K to the set of all wffs such that the following
conditions (clauses) are satisfied for every propositional variable p, wffs A,
B and a ∈ K:
(i). (a ≤ b & a p) ⇒ b p
(ii). a A ∧ B iff a A and a B
(iii). a A ∨ B iff a A or a B
(iv). a A → B iff for all b, c ∈ K, (Rabc and b A) ⇒ c B
(v). a ¬A iff a∗ 2 A
Definition 2.10 (Truth in a BRM3-model). A wff A is true in a BRM3model iff a A for all a ∈ O in this model.
Definition 2.11 (BRM3-validity). A formula A is BRM3-valid (in symbols, BRM3 A) iff a A for all a ∈ O in all BRM3-models.
Remark 2 (B-models). A B-model, that is, a model for Routley and
Meyer’s basic logic B (Definition 2.5) is defined similarly as a BRM3-model,
except that postulates P6, P7 and P8 are dropped (cf. [14], Chapter 4).
112
GEMMA ROBLES
Models for logics extending BRM3 are simply defined by adding to P1P8 the appropriate semantical postulates while defining ‘truth in a model’
and ‘validity’ similarly as in Definitions 2.10 and 2.11. In this way, an RMsemantics for a wealth of extensions of BRM3 (RM3 included) is defined in
Section 4.
In order to prove soundness, the following lemmas are useful. (An
adequate version of these lemmas is immediate for each extension of BRM3
considered in this paper.)
Lemma 2.12 (Hereditary condition). For any BRM3-model, a, b ∈ K
and wff A, (a ≤ b & a A) ⇒ b A.
Proof. Induction on the length of A. The conditional case is proved
with P2 and the negation case with P5.
Lemma 2.13 (Entailment lemma). For any wffs A, B, BRM3 A → B
iff (a A ⇒ a B, for all a ∈ K) in all BRM3-models.
Proof. From left to right: by P1; from right to left: by Lemma 2.12.
We can now prove soundness.
Theorem 2.14 (Soundness of BRM3). For each wff A, if `BRM3 A,
then BRM3 A.
Proof. Axioms A1-A8 and the rules MP, Adj, Suf, Pref and Con are
proved as in B-models (cf. Remark 2; cf. [14]). Then, it remains to
prove that A9, A10 and A11 are BRM3-valid. We proceed by ‘reductio ad
absurdum’ (by clauses ii-v, we refer to those in Definition 2.9; the proofs
are simplified by leaning on Lemma 2.13).
(a) A9 A → (A → A) is BRM3-valid: Suppose that there is a ∈ K
in some BRM3-model and wff A such that (1) a A but a 2 A → A.
Then, (2) b A, c 2 A for b, c ∈ K such that Rabc (clause iv). By P6, (3)
a ≤ c or b ≤ c. By applying Lemma 2.12 to 1, 2 and 3, we get (4) c A,
contradicting 2.
(b) A10 (A ∧ ¬A) → (B ∨ ¬B) is BRM3-valid: Suppose that there is
a ∈ K in some BRM3-model and wffs A, B such that (1) a A ∧ ¬A but
(2) a 2 B ∨ ¬B. Then, (3) a A, a∗ 2 A (i.e., a ¬A), a 2 B, a∗ B
(i.e., a 2 ¬B) by clauses ii, iii and v. By P7, (4) a ≤ a∗ or a∗ ≤ a. So, (5)
either a∗ A or a B (3, 4, Lemma 2.12), a contradiction.
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
113
(c) A11 ¬A → [A ∨ (A → B)] is BRM3-valid: Suppose that there is
a ∈ K in some BRM3-model and wffs A, B such that (1) a ¬A but
a 2 A ∨ (A → B), i.e., a 2 A and a 2 A → B (clause iii). Then, (2) a∗ 2 A,
a 2 A, b A, c 2 B for b, c ∈ K such that Rabc (clauses iv, v). By P8, (3)
b ≤ a or (4) b ≤ a∗ . But if 3 is the case, b 2 A follows (2 and Lemma 2.12),
contradicting 2; and if 4 is the case, then we have a∗ A (2 and Lemma
2.12), also contradicting 2.
In the next section, we prove the completeness of BRM3 w.r.t. the
RM-semantics for this logic developed in the present section.
3. Completeness of BRM3
We begin by defining some preliminary concepts necessary in order to define
the canonical model (cf. [14], Chapter 4).
Definition 3.1 (BRM3-theories). A BMR3-theory (theory, for short)
is a set of formulas closed under Adjunction (Adj) and BRM3-implication
(BRM3-imp). That is, a is a theory if whenever A, B ∈ a, A ∧ B ∈ a; and
if whenever A → B is a theorem of BRM3 and A ∈ a, B ∈ a.
Definition 3.2 (Classes of theories). Let a be a theory. We set (1) a
is prime iff whenever A ∨ B ∈ a, then A ∈ a or B ∈ a; (2) a is empty iff it
contains no wffs; (3) a is regular iff a contains all theorems of BRM3; (4)
a is trivial iff every wff belongs to it.
Next, the canonical model is defined.
Definition 3.3 (The canonical BRM3-model). Let K T be the set of
all theories and RT be defined on K T as follows: for all a, b, c ∈ K T and
wffs A, B, RT abc iff (A → B ∈ a & A ∈ b) ⇒ B ∈ c. Now, let K C be
the set of all non-trivial, non-empty prime theories and OC be the subset
of K C formed by the regular theories. On the other hand, let RC be the
restriction of RT to K C and ∗C be defined on K C as follows: for each
a ∈ K C , a∗ = {A | ¬A ∈
/ a}. Finally, C is defined as follows: for any
a ∈ K C and wff A, a C A iff A ∈ a. Then, the canonical model is the
structure (K C , OC , RC , ∗C , C ).
114
GEMMA ROBLES
The canonical model for any extension S of BRM3 is defined in a similar
way but referring its items to S-theories.
Before proceeding into the completeness proof, let us note an important
fact.
Remark 3 (On canonical relevant models). There is an important feature distinguishing the canonical BRM3-model (and those for the extensions
of BRM3 considered in this paper) from the canonical models for standard
relevant logics such as E (the logic of entailment) or R (the logic of relevant
conditional). Namely, in the former, theories need to be non-empty and
non-trivial, unlike in the latter. This fact permeates the entire completeness proof sharply distinguishing it from standard proofs for E, R or their
subsystems: each time a theory is built, one has to show that it contains
(and lacks) at least one wff. In this sense, A9 and A11 are essential. On the
other hand, the canonical model for BRM3 (and those for its extensions) is
differentiated from those (in the RM-semantics) for Lukasiewicz’s 3-valued
logic L3 (cf [11]) or 3-valued logic G3L (cf. [9]), say, in the following respect: unlike in BRM3 and its extensions included in RM3, non-triviality
is equivalent to weak consistency (absence of the negation of any theorem;
cf. [10] on this notion )in L3 and G3L . Actually, as shown in Proposition
A6, there are regular, non-trivial, prime theories containing the negation
of a theorem.
We proceed into the completeness proof. Firstly, a series of lemmas
is proved leaning on which it will be shown that the structure defined in
Definition 3.3 is indeed a BRM3-model.
Lemma 3.4 (Defining x for a, b in RT ). Let a, b be non-empty theories.
The set x = {B | ∃A[A → B ∈ a & A ∈ b]} is a non-empty theory such
that RT abx.
Proof. Assume the hypothesis of Lemma 3.4 and define x as indicated.
It is easy to show that x is a theory. Then, RT abx is immediate by definition
of RT (Definition 3.3). Moreover, x is non-empty. Let A ∈ a, B ∈ b. By
T4, A → [B → (A ∨ B)]. So, B → (A ∨ B) ∈ a and thus, A ∨ B ∈ x.
Lemma 3.5 (Extending a in RT abc to a member in K C ). Let a, b be
non-empty theories and c be a non-trivial prime theory such that RT abc.
Then, there is a non trivial (and non-empty) prime theory x such that a ⊆ x
and RT xbc.
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
115
Proof. Given the hypothesis of Lemma 3.5, we can build a non-empty
prime theory x such that a ⊆ x and RT xbc, following [14], Chapter 4.
(The proof works for almost any logic including B+ —cf. Definition 2.5.)
Suppose now that x is trivial and let A ∈ b and B be an arbitrary wff. As
x is trivial, A → B ∈ x. Then, B ∈ c (RT xbc, A → B ∈ x, A ∈ b and
definition of RT —cf. Definition 3.3), contradicting the non-triviality of c.
Lemma 3.6 (Extending b in RT abc to a member in K C ). Let a and b
be non-empty theories and c be a non-trivial prime theory such that RT abc.
Then, there is a non trivial (and non-empty) prime theory x such that b ⊆ x
and RT axc.
Proof. Similarly as in the preceding lemma, we build a non-empty
prime theory x such that RT axc. Suppose that x is trivial and let A ∈ a
and B be an arbitrary wff. By T5, A → [¬A ∨ (¬A → B)]. So, ¬A ∨ (¬A →
B) ∈ a, whence, by primeness of a, either (1) ¬A ∈ a or (2) ¬A → B ∈ a.
Let us first consider case 2. As x is trivial, ¬A ∈ x. But then B ∈ c
(RT axc, ¬A → B ∈ a, ¬A ∈ x and definition of RT ), contradicting the aconsistency of c. Let us now examine case 1. Firstly, notice that A∧¬A ∈ a.
Then, B ∨ ¬B ∈ a, by applying A10 (A ∧ ¬A) → (B ∨ ¬B). By hypothesis,
B ∈
/ a, so ¬B ∈ a. By A11,¬B → [B ∨ (B → B)] is a theorem. Thus,
B ∨ (B → B)] ∈ a, that is, B → B ∈ a because B ∈
/ a. But B ∈ x for x
T
is trivial. Then, B ∈ c (R axc, B → B ∈ a, B ∈ x and definition of RT ),
contradicting the a-consistency of c.
Lemma 3.7 below shows that the canonical relation ≤C is just set inclusion between non-trivial and non-empty prime theories.
Lemma 3.7 (≤C and ⊆ are coextensive). For any a, b ∈ K C , a ≤C b
iff a ⊆ b.
Proof. From left to right, it is immediate. So, suppose a ⊆ b for
a, b ∈ K C . Clearly, RT BRM3aa (cf. Definition 3.1 and Definition 3.3).
Then, by using Lemma 3.5, there is some regular non-trivial theory x such
that BRM3 ⊆ x and RC xaa. By the hypothesis RC xab, i.e., a ≤C b, since
x ∈ OC .
Lemma 3.8 (Primeness of ∗-images). Let a be prime theory. Then,
(1) a∗ is a prime theory as well; (2) for any wff A, ¬A ∈ a∗ iff A ∈
/ a.
116
GEMMA ROBLES
Proof. (Cf. [14], Chapter 4.) (1) a∗ is closed under BRM3-imp by
Con; a∗ is closed under Adj by T3; a∗ is prime by T2. (2) By A7 and A8.
Lemma 3.9 (∗C is an operation on K C ). Let a be a non-trivial and
non-empty prime theory. Then, a∗ is a non-trivial and non-empty prime
theory as well.
Proof. By Lemma 3.8(1), a∗ is a prime theory. Next, it is shown that
if a is non-trivial and non-empty, then a∗ is also non-trivial and non-empty.
(1) a∗ is non-empty: as a is non-trivial, there is some wff A such that A ∈
/ a.
Then, ¬A ∈ a∗ by Lemma 3.8(2). (2) a∗ is non-trivial. As a is non-empty,
there is some wff A such that A ∈ a. Then, ¬A ∈
/ a∗ by Lemma 3.8(2). Concerning Lemma 3.9, we note the following remark.
Remark 4 (∗C is not an operation on OC ). The canonical operation
∗C is not an operation on OC : a would have to be weak consistent in order
to prove that a∗ is regular (cf. Remark 3).
In what follows, we prove the following two facts: (1) postulates P1-P8
hold in the canonical model; and (2) C is a (valuation) relation satisfying
clauses i-v in Definition 3.3.
Lemma 3.10 (P1-P8 hold canonically). Postulates P1-P8 hold in the
canonical BRM3-model.
Proof. The use of Lemma 3.7 will greatly simplify the proof. By
leaning on this lemma, P1-P5 are proved similarly as in [14], Chapter 4.
So, let us prove P6, P7 and P8. We proceed by ‘reductio ad absurdum’.
P6 (RC abc ⇒ (a ≤C c or b ≤C c)) holds in the canonical BRM3-model:
suppose that there are a, b, c ∈ K C and wffs A, B such that (1) RC abc but
(2) A ∈ a, A ∈
/ c, B ∈ b and B ∈
/ c. By T4, (3) A → [B → (A ∨ B)]. So,
we have (4) B → (A ∨ B) ∈ a (by 2, 3) and (5) A ∨ B ∈ c (by 1, 2, 4). But
5 contradicts 2.
P7 (a ≤ a∗ or a∗ ≤ a) holds in the canonical RM3-model: suppose there
is a ∈ K C and wffs A, B such that (1) A ∈ a, A ∈
/ a∗ (i.e., ¬A ∈ a), B ∈ a∗
(i.e., ¬B ∈
/ a) and B ∈
/ a. By A10, (2) (A ∧ ¬A) → (B ∨ ¬B). So, we have
(3) B ∨ ¬B ∈ a (by 1). But 3 contradicts 1.
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
117
P8 (RC abc ⇒ (b ≤C a or b ≤C a∗ )) holds in the canonical BRM3–
model: suppose that there are a, b, c ∈ K C and wffs A, B such that (1)
RC abc but (2) A ∈ b, A ∈
/ a, B ∈ b and B ∈
/ a∗ (i.e., ¬B ∈ a). By A11,
(3) ¬B → [B ∨ (B → C)] for arbitrary wff C. Then, (4) B ∨ (B → C) ∈ a
(by 2, 3) whence (5) B ∈ a or (6) B → C ∈ a. Let us consider the second
alternative, 6. By applying the definition of RC to 1, 2 (B ∈ b) and 6, we
have (7) C ∈ c, contradicting the non-triviality of c. So, let us consider the
first alternative, 5. We have (8) B∧¬B ∈ a (by 2, 5), whence (9) A∨¬A ∈ a
follows by A10 (B ∧ ¬B) → (A ∨ ¬A). Now, (10) ¬A ∈ a for A ∈
/ a by 2.
Next, again by A11, we have (11) ¬A → [A ∨ (A → C)] for arbitrary wff C.
So, (12) A ∨ (A → C) ∈ a and, since A ∈
/ a, then (13) A → C ∈ a. Finally,
by 1, 2 and (13), C ∈ c, contradicting the non-triviality of c.
Before showing that the clauses hold canonically, we prove the primeness
lemma.
Lemma 3.11 (Extension to prime theories). Let a be a theory and A
a wff such that A ∈
/ a. Then, there is a prime theory x such that a ⊆ x and
A∈
/ x.
Proof. Cf. [14], Chapter 4, where it is shown how to proceed in an
ample class of logics including the logic B+ (cf. Definition 2.5).
Lemma 3.12 (Clauses i-v hold canonically). Clauses i-v in Definition
2.9 are satisfied by the canonical BRM3-model.
Proof. Clause i is immediate by Lemma 3.7 and clauses ii, iii and iv
from left to right are very easy. So, let us prove iv from right to left. For
wffs A, B and a ∈ K C , suppose A → B ∈
/ a (i.e., a 2C A → B). We
C
prove that there are x, y ∈ K such that RC axy, A ∈ x (i.e., x C A)
and B ∈
/ y (i.e., y 2C B). Consider the sets z = {C |`BRM3 A → C} and
u = {C | ∃D[D → C ∈ a & D ∈ z]}. They are theories such that RT azu.
Now, A ∈ z (by A1) and B ∈
/ u (if B ∈ u, then A → B ∈ a, contradicting
the hypothesis). So, z is non-empty and u is non-trivial. Moreover, u is
non-empty by Lemma 3.4. Now, by applying Lemma 3.11, u is extended to
a non-trivial, non-empty prime theory y such that u ⊆ y, B ∈
/ y and RT azy.
Next, by using Lemma 3.6, z is extended to a non-trivial, non-empty prime
theory x such that z ⊆ x and RC axy. Clearly, A ∈ x. Therefore, we have
118
GEMMA ROBLES
non-trivial and non-empty prime theories x, y such that A ∈ x, B ∈
/ y and
RC axy, as was to be proved.
After showing that the canonical model is indeed a model, we finally
prove completeness.
Lemma 3.13 (The canonical model is in fact a model). The canonical
BRM3-model is in fact a BRM3-model.
Proof. Since RC is clearly a ternary relation on K C , ∗C is an operation
on K C (Lemma 3.9) and K C is non-empty (Lemma 3.11: BRM3 is nonempty and non-trivial), Lemma 3.13 follows by Lemma 3.10 and 3.12. Theorem 3.14 (Completeness of BRM3). For each wff A, if BRM3 A,
then `BRM3 A.
Proof. Suppose 0BRM3 A. By Lemma 3.11, there is a non-trivial, nonempty prime theory x such that BRM3 ⊆ x and A ∈
/ x. By Definition 3.3
C
and Lemma 3.13, x 2 A. Therefore, 2BRM3 A by Definition 2.11.
4. Extensions of BRM3 included in RM3
In this section, we provide an RM-semantics for a wealth of extensions of
BRM3 included in the logic RM3 among which RM3 itself is to be found.
Consider the theses recorded in the following proposition.
Proposition 4.1 (Theses provable in RM3). The theses and the rule
that follow are derivable in RM3. (In some cases some equivalent (w.r.t.
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
119
BRM3) variants are remarked):
t1. (A → B) → [(B → C) → (A → C)]
t2. (B → C) → [(A → B) → (A → C)]
t3. [(A → B) ∧ (B → C)] → (A → C)
t4. [A ∧ (A → B)] → B ( = ¬B → [¬A ∨ ¬(A → B)])
t5. [A → (A → B)] → (A → B) ( = [A → (B → C)] → [(A ∧ B) → C])
t6. [A → (B → C)] → [(A → B) → (A → C)]
t7. (A → B) → [[A → (B → C)] → (A → C)]
t8. A ⇒ (A → B) → B
t9. (A → B) → [[A ∧ (B → C)] → C]
t10. (A ∧ B) → [[A → (B → C)] → C] ( = A → [[A → (A → B)] → B])
t11. [A ∧ (B → C)] → [(A → B) → C)]
t12. A → [(A → B) → B]
t13. [A → (B → C)] → [B → (A → C)]
t14. B → [[A → (B → C)] → (A → C)]
t15. (A → B) ∨ (B → A)
t16. [(A ∧ B) → C] → [(A → C) ∨ (B → C)]
t17. [A → (B ∨ C)] → [(A → B) ∨ (A → C)]
t18. A ∨ (A → B)
t19. (A → B) → (¬B → ¬A) ( = (A → ¬B) → (B → ¬A) =
(¬A → B) → (¬B → A) = (¬A → ¬B) → (B → A))
t20. ¬(A ∧ ¬A) ( = A ∨ ¬A)
t21. (A ∧ ¬B) → ¬(A → B) ( = (A → B) → (¬A ∨ B))
t22. [(A → B) ∧ ¬B] → ¬A ( = A → [B ∨ ¬(A → B)])
t23. (¬A ∧ B) → (A → B) ( = ¬(A → B) → (A ∨ ¬B))
t24. ¬(A → B) → (B → A)
t25. ¬A → [¬B ∨ (A → B)]
t26. (A ∨ ¬B) ∨ (A → B)
t27. ¬A ∨ (B → A)
t28. A ∨ [A → (B ∨ ¬B)]
t29. [¬(A → B) ∧ B] → ¬B ( = B → [¬B ∨ (A → B)])
t30. B → [(A ∧ ¬A) → B]
120
GEMMA ROBLES
Proof. By the matrix MRM3 (cf. Definition 2.4): t8 preserves RM3–
validity and t1-t7, t9-t30 are RM3-valid.
On the other hand, we have:
Proposition 4.2 (t1-t30 are not provable in BRM3). Theses t1-t30 are
not provable in BRM3.
Proof. See the appendix.
Let now S be an extension of BRM3 axiomatized by any selection of
t1-t30. The aim of this section is to define an RM-semantics for S. The key
concept is “corresponding postulate (cp) to a thesis or rule”, which can be
rendered as follows (cf. [14], p. 301).
Definition 4.3 (Corresponding postulate —cp). Let ti be any of t1t30, and let pj be a semantical postulate. Then, given the logic BRM3
and BRM3-models, pj is the cp to ti iff (1) ti is true in any BRM3-model
in which pj holds; and (2) pj holds in the canonical BRM3-model if ti is
added as an axiom (or rule) to BRM3.
It must be clear that if, given the logic BRM3 and BRM3-semantics,
pj is the cp to ti, then the logic BRM3 + ti (i.e., BRM3 plus ti) is sound
and complete w.r.t. BRM3-models (i.e., BRM3-models where pj holds).
Given a BRM3-model M, consider the following definition and semantical postulates for all a, b, c, d ∈ K with quantifiers ranging over K (in some
cases, some equivalent (w.r.t. BRM3-models) variants are remarked):
d2. R2 abcd =df ∃x(Rabx & Rxcd)
Pt1. R2 abcd ⇒ ∃x(Racx & Rbxd)
Pt2. R2 abcd ⇒ ∃x(Rbcx & Raxd)
Pt3. Rabc ⇒ ∃x(Rabx & Raxc)
Pt4. Raaa ( = Ra∗ a∗ a∗ )
Pt5. Rabc ⇒ R2 abbc
Pt6. R2 abcd ⇒ ∃x, y(Racx & Rbcy & Rxyd)
Pt7. R2 abcd ⇒ ∃x, y(Racx & Rbcy & Ryxd)
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
121
Pt8. (∃x ∈ O) Raxa
Pt9. Rabc ⇒ ∃x(Rabx & Rbxc)
Pt10. Rabc ⇒ R2 baac
Pt11. Rabc ⇒ ∃x(Rbax & Raxc)
Pt12. Rabc ⇒ Rbac
Pt13. R2 abcd ⇒ R2 acbd
Pt14. R2 abcd ⇒ R2 bcad
Pt15. (a ∈ O & Rabc & Rade) ⇒ (b ≤ e or d ≤ c)
Pt16. (Rabc & Rade) ⇒ ∃x[b ≤ x & d ≤ x & (Raxc or Raxe)]
Pt17. (Rabc & Rade) ⇒ ∃x[x ≤ c & x ≤ e & (Rabx or Radx)]
Pt18. (a ∈ O & Rabc) ⇒ b ≤ a
Pt19. Rabc ⇒ Rac∗ b∗
Pt20. a ∈ O ⇒ a ≤ a∗ ( = a ∈ O ⇒ a∗ ≤ a∗∗ )
Pt21. Raa∗ a ( = Ra∗ aa∗ )
Pt22. Ra∗ aa ( = Raa∗ a∗ )
Pt23. Rabc ⇒ (a ≤ c or b ≤ a∗ )
Pt24. (Rabc & Ra∗ de) ⇒ (b ≤ e or d ≤ c)]
Pt25. Rabc ⇒ (b ≤ a∗ or a∗ ≤ c)
Pt26. (a ∈ O & Rabc) ⇒ (a∗ ≤ c or b ≤ a)
Pt27. (a ∈ O & Rabc) ⇒ a∗ ≤ c
Pt28. Rabc ⇒ (b ≤ a or c∗ ≤ c)
Pt29. Ra∗ bc ⇒ (a ≤ c or a∗ ≤ c)
Pt30. Rabc ⇒ (a ≤ c or b ≤ b∗ )
We have the following proposition:
Proposition 4.4 (Corresponding postulates to t1-t30). Given the logic
BRM3 and BRM3-models, ptk is the corresponding postulate (cp) to tk
(1 ≤ k ≤ 30).
Proof. The proof is similar to that provided in [14], Chapter 4, for
extensions of Routley and Meyer’s basic logic B (cf. Definition 2.5). Actually, t1-t8, t12, t13, t15, t16, t18, t19-t21 and t24 are among the theses and
122
GEMMA ROBLES
rules considered in [14]. However, notice that, as it was remarked above
(cf. Remark 3), any new theory introduced has to be shown non-empty and
non-trivial. Consequently, most of the proofs in [14] have to modified in the
context of the present paper. Nevertheless, the required modifications are
easy to get by using Lemmas 3.4-3.7, as actually shown above in the case
of postulates P6, P7 and P8. Anyway, let us prove a couple of examples
(we follow the pattern set on in Theorem 2.14 and Lemma 3.10).
(a) pt11 is the cp to t11. (ai) t11 is true in any BRM3 + pt11-model :
suppose that there is a ∈ K in some BRM3 + pt11-model and wffs A, B
such that (1) a A ∧ (B → C) but (2) a 2 (A → B) → C. Then, (3)
a A, a B → C (clause ii, 1) and (4) b A → B, c 2 C for b, c ∈ K
such that Rabc in this model (clause iv, 2). By pt11, we have (5) Rbax
and (6) Raxc for some x ∈ K. Then, we get (7) x B (3, 4, 5, clause iv)
and, finally, (8) c C (3, 6, 7, clause iv), contradicting 4. (aii) pt11 holds
in the canonical BRM3 + t11-model : for a, b, c ∈ K C , suppose (1) RC abc.
Consider now the following set y = {B | ∃A[A → B ∈ b & A ∈ a]}. By
Lemma 3.4, y is a non-empty theory such that (2) RT bay. Next, we show
that RT ayc holds. Suppose (3) A → B ∈ a and (4) A ∈ y for wffs A, B.
We have to show that B ∈ c. By definition of y, we have (5) C → A ∈ b for
some wff C such that (6) C ∈ a. By t11, [C ∧ (A → B)] → [(C → A) → B]
is a theorem. So, (7) (C → A) → B ∈ a since C ∧ (A → B) ∈ a (by 3, 6).
Then, (8) B ∈ c (by 1, 5, 7). Thus, we have a non-empty theory y such
that (9) RT bay and RT ayc. By Lemma 3.6, y is extended to a non-trivial,
non-empty prime theory x such that y ⊆ x and RC axc. Obviously, RC bax.
Therefore, we have x ∈ K C such that RC bax and RC axc, as it was to be
proved.
(b) pt26 is the cp to t26. (bi) t26 is true in any BRM3 + pt26-model :
suppose that there is a ∈ O in some BRM3 + pt26-model and wffs A, B
such that (1) a 2 (A ∨ ¬B) ∨ (A → B), i.e., (2) a 2 A, (3) a∗ B (i.e.,
a 2 ¬B by clause v) and (4) a 2 A → B by applying clause iii to 1. Then,
(5) b A and c 2 B for b, c ∈ K such that Rabc in this model (by clause
iv in 4). By pt26, (6) a∗ ≤ c or (7) b ≤ a. Suppose 6. Then, (8) c B (by
3), contradicting 5. On the other hand, suppose 7. Then, (9) a A (by
5), contradicting 2. (bii) pt26 holds in the canonical BRM3 + t26-model :
suppose that there are a ∈ OC and b, c ∈ K C and wffs A, B, C such that
(1) RC abc but (2) A ∈ a∗ (i.e., ¬A ∈
/ a by clause v), A ∈
/ c, B ∈ b and
B∈
/ a. By t26, (3) (B ∨ ¬A) ∨ (B → A) ∈ a. So, (4) B → A ∈ a follows by
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
2 and 3. Finally, (5) A ∈ c (by 1, 2 and 4), contradicting 2.
123
The section is ended by defining a general Routley-Meyer semantics for
RM3.
As pointed out in the appendix, following Anderson and Belnap, RM3
can be axiomatized by adding A9 (A → (A → A)) and t18 (A ∨ (A → B))
to the logic of relevance R (cf. [1], pp. 469, ff.). Now, a model for R can
be defined as follows (cf. [14], Chapter 4).
Definition 4.5 (R-models). An R-model is structure (K, O, R, ∗, )
where K, O, R, ∗ and are defined similarly as in B-models (cf. Remark 2),
save for the addition of the following semantical postulates: pt1 (R2 abcd ⇒
∃x(Racx & Rbxd)), pt5 (Rabc ⇒ R2 abbc), pt12 (Rabc ⇒ Rbac) and pt19
(Rabc ⇒ Rac∗ b∗ ).
Then, an RM3-model is defined as follows:
Definition 4.6 (RM3-models 1). An RM3-model is defined similarly
as an R-model, save for the addition of postulates: pt6 (Rabc ⇒ (a ≤ c or
b ≤ c)) and pt18 ((a ∈ O & Rabc) ⇒ b ≤ a).
But, on the other hand, following Brady, RM3 can be axiomatized as an
extension of the logic DW with the following axioms: A11, t4, t21 and t23
(cf. [2]; cf. the appendix). So, RM3-models can alternatively be defined as
follows (firstly, DW-models are defined).
Definition 4.7 (DW-models). A DW-model is defined similarly as a Bmodel (cf. Remark 2), except that P5 is changed for pt19 (Rabc ⇒ Rac∗ b∗ ).
Definition 4.8 (RM3-models 2). An RM3-model is defined similarly as
a DW-model, save for the addition of the following semantical postulates:
P8 (Rabc ⇒ (b ≤ a or b ≤ a∗ )), pt4 (Raaa), pt21 (Raa∗ a) and pt23
(Rabc ⇒ (a ≤ c or b ≤ a∗ )).
Notice that in Definition 4.5 and 4.6, postulate P5 is derivable immediately by pt19. Also, remark that P7 is not necessary in Definitions 4.6
and 4.8, since A10 is, of course, derivable in RM3 and P7 is the cp to A10
w.r.t. B-models (cf. Remark 2 and Lemma 3.10).
124
GEMMA ROBLES
5. Conclusions
In this paper, we have provided an RM-semantics for an ample family of
sublogics of RM3 including the basic logic BRM3. Actually, in Section 4,
it has been shown that each one of t1-t30 can be added to BMR3 independently, the resulting logic being characterized by an RM-semantics. And,
as pointed out in the introduction to this paper, these logics can be useful when “some relevance”, but not the full vsp, is needed. Nonetheless,
we have not paused to discuss neither BRM3 nor any of its extensions.
So, future work on the topic could focus on this question or on proposing
alternative extensions of BRM3 not defined in this paper. On our part,
we will close these brief considerations by remarking that the matrices in
section 2 in the appendix provide a rough first selection of sublogics of
RM3. Thus, for example, M3 shows (when 1 is designated in addition to
2) that the characteristic axioms of RM3 (A9 and A11 together with A10)
can be added to ticket entailment logic, T (minus the reductio axiom t21,
but with the ‘Principium of tertium non datur’, t20), without the result
collapsing in RM3 (T is axiomatized when dropping t12 and adding t2 to
the formulation of R in Definition A3 in the appendix).
A.
. Appendix
A.1. Independence in BRM3
The following matrices show that A7, A8, A9, A10, A11 and Con are independent of each other, given the logic B+ (cf. Proposition 2.5 on the logic
B+ ; cf. Definitions 2.2, 2.3 on the notion of a logical matrix. Designated
values are starred):
Matrix I. Independence of A7:
→
0
*1
0
1
0
1
1
1
¬
0
0
∧
0
*1
0
0
0
1
0
1
Falsifies A7 (A = 1).
Matrix II. Independence of A8:
∨
0
*1
0
0
1
1
1
1
125
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
Same set as in matrix I (the positive classical truth tables), save for the
table for negation which is now as follows:
0
*1
¬
1
1
Falsifies A8 (A = 0).
Matrix III. Independence of Con:
→
0
1
2
*3
*4
0
4
2
1
0
0
1
4
3
1
1
0
2
4
2
3
2
0
3
4
3
3
3
0
4
4
4
4
4
4
¬
4
2
1
3
0
∧
0
1
2
*3
*4
0
0
0
0
0
0
1
0
1
0
1
1
2
0
0
2
2
2
3
0
1
2
3
3
4
0
1
2
3
4
∨
0
*1
*2
0
0
1
2
∨
0
1
2
*3
*4
0
0
1
2
3
4
1
1
1
3
3
4
2
2
3
2
3
4
3
3
3
3
3
4
4
4
4
4
4
4
Falsifies Con (A = 1, B = 3).
Matrix IV. Independence of A9:
→
0
*1
*2
0
2
0
0
1
2
1
0
2
2
1
1
¬
2
1
0
∧
0
*1
*2
0
0
0
0
1
0
1
1
2
0
1
2
1
1
1
2
2
2
2
2
∨
0
1
*2
*3
0
0
1
2
3
Falsifies A9 (A = 2).
Matrix V. Independence of A10:
→
0
1
*2
*3
0
3
0
0
0
1
3
3
0
0
2
3
0
2
0
3
3
3
3
3
¬
3
1
2
0
∧
0
1
*2
*3
0
0
0
0
0
1
0
1
0
1
2
0
0
2
2
3
0
1
2
3
1
1
1
3
3
2
2
3
2
3
3
3
3
3
3
Falsifies A10 (A = 2, B = 1).
Matrix VI. Independence of A11:
The tables for ∧, ∨, ¬ are as in Matrix IV, but the conditional table is
as follows:
126
GEMMA ROBLES
→
0
*1
*2
0
1
0
0
1
1
1
0
2
2
2
2
Falsifies A11 (A = B = 0).
A.2. t1-t30 are not provable in BRM3
We prove that t1-t30 are not derivable in BRM3. We shall use eight
matrices. Each one of them verifies BRM3 and falsifies some subset of
t1-t30 (designated values are starred).
Matrix M1:
The tables for ∧, ∨, ¬ are as in Matrix IV in the preceding section, but
the conditional table is as follows:
→
0
1
*2
0
2
1
0
1
2
2
0
2
2
2
2
Falsifies t1 (A = 1, B = C = 0); t2 (A = B = 1, C = 0); t4 (A = 1,
B = 0); t5 (A = 1, B = 0); t6 (A = B = 1, C = 0); t7 (A = B = 1, C = 0);
t9 (A = B = 1, C = 0); t19 (A = 1, B = 0); t20 (A = 1); t27 (A = 1,
B = 2).
Matrix M2:
The tables for ∧, ∨, ¬ are as in M1, but the conditional table is the
following:
→
0
1
*2
0
2
1
0
1
2
2
1
2
2
2
2
Falsifies t3 (A = 2, B = 1, C = 0).
Matrix M3:
The tables for ∧, ∨, ¬ are as in M1, but the conditional table is as
follows:
127
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
→
0
*1
*2
0
2
0
0
1
2
2
0
2
2
2
2
Falsifies t8 (A = B = 1); t10 (A = B = C = 1); t11 (A = B = C = 1);
t12 (A = B = 1); t13 (A = 2, B = C = 1); t14 (A = 2, B = C = 1); t18
(A = 1, B = 0); t21 (A = B = 1).
Matrix M4:
The tables for ∧, ∨, ¬ are as in M1, but the conditional table is the
following:
→
0
1
*2
0
2
0
0
1
2
2
1
2
2
2
2
Falsifies t22 (A = 2, B = 1).
Matrix M5:
→
0
1
2
*3
0
3
2
0
0
1
3
3
0
0
2
3
3
3
0
3
3
3
3
3
¬
3
2
1
0
∧
0
1
2
*3
0
0
0
0
0
1
0
1
1
1
2
0
1
2
2
∨
0
1
2
*3
3
0
1
2
3
0
0
1
2
3
1
1
1
2
3
2
2
2
2
3
3
3
3
3
3
Falsifies t23 (A = 2, B = 1); t26 (A = 2, B = 1); t29 (A = 3, B = 2).
Matrix M6:
→
0
1
2
3
*4
0
4
3
0
0
0
1
4
4
0
0
0
2
4
4
4
0
0
3
4
4
4
4
0
4
4
4
4
4
4
¬
4
3
2
1
0
∧
0
1
2
3
*4
0
0
0
0
0
0
1
0
1
1
1
1
2
0
1
2
2
2
3
0
1
2
3
3
4
0
1
2
3
4
Falsifies t28 (A = 3, B = 2); t30(A = 2, B = 1).
∨
0
1
2
3
*4
0
0
1
2
3
4
1
1
1
2
3
4
2
2
2
2
3
4
3
3
3
3
3
4
4
4
4
4
4
4
128
GEMMA ROBLES
Matrix M7:
→ 0 1
0
5 5
4 5
1
2
3 3
3
2 2
1 1
4
*5 0 0
∨
0
1
2
3
4
*5
0
0
1
2
3
4
5
1
1
1
2
3
4
5
2
5
5
5
2
2
0
3
5
5
3
5
3
0
4
5
5
5
5
5
0
5
5
5
5
5
5
5
2
2
2
2
4
4
5
3
3
3
4
3
4
5
4
4
4
4
4
4
5
5
5
5
5
5
5
5
∧
0
1
2
3
4
*5
¬
5
4
3
2
1
0
0
0
0
0
0
0
0
1
0
1
1
1
1
1
2
0
1
2
1
2
2
3
0
1
1
3
3
3
4
0
1
2
3
4
4
5
0
1
2
3
4
5
Falsifies t15 (A = 3, B = 2); t16 (A = 3, B = 2, C = 1); t17 (A = 4,
B = 3, C = 2).
Matrix M8:
→ 0 1
0
6 6
1
5 6
2
4 4
3
3 4
4
0 0
5
0 0
*6 0 0
∨
0
1
2
3
4
5
*6
0
0
1
2
3
4
5
6
1
1
1
3
3
4
5
6
2
6
5
6
5
0
0
0
3
6
6
6
6
0
0
0
4
6
6
6
6
6
0
0
5
6
6
6
6
0
6
0
6
6
6
6
6
6
6
6
2
2
3
2
3
4
5
6
3
3
3
3
3
4
5
6
4
4
4
4
4
4
6
6
5
5
5
5
5
6
5
6
6
6
6
6
6
6
6
6
¬
6
5
4
3
2
1
0
∧
0
1
2
3
4
5
*6
0
0
0
0
0
0
0
0
1
0
1
0
1
1
1
1
Falsifies t24 (A = 5, B = 4); t25 (A = 5, B = 4).
2
0
0
2
2
2
2
2
3
0
1
2
3
3
3
3
4
0
1
2
3
4
3
4
5
0
1
2
3
3
5
5
6
0
1
2
3
4
5
6
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
129
A.3. Axiomatization of RM3
We record the formulations of RM3 provided in [1] and in [2], respectively.
Definition A.1 (The logic DW). The logic DW is axiomatized when
changing the rule Con for t19 in the formulation of Routley and Meyer’s
logic B (cf. Definition 2.5; Proposition 4.1; cf. [6] about DW and other
weak relevant logics).
Definition A.2 (Brady’s axiomatization of RM3). RM3 can be axiomatized by adding to DW the following axioms: A11, t4, t21 and t23 (cf. [2];
cf. Proposition 4.1).
Definition A.3 (The logic R). Anderson and Belnap’s logic of relevance R can be axiomatized with the following axioms and rules: A1, A2,
A3, A4, A5, A6, A7, A8, t1, t5, t12, t19, MP and Adj (cf. [1]; Definition
2.5; Proposition 4.1).
Definition A.4 (Anderson and Belnap’s axiomatization of RM3). According to Anderson and Belnap (cf. [1], p. 469, ff.), RM3 can be axiomatized by adding to R the following axioms: t18 (A ∨ (A → B)) and
A → (¬A → A).
Now, A → (¬A → A) and A9 (A → (A → A)) are easily shown
equivalent, given R. So, RM3 can be axiomatized by adding A9 and t18 to
R. Moreover, since t18 is immediate given R and A11 (¬A → [A ∨ (A →
B)]), RM3 can be axiomatized by adding A9 and A11 to R.
A.4. RM3 has the quasi-relevance property
We prove:
Proposition A.5 (RM3 is quasi-relevant). If A → B is a theorem of
RM3, then either (1) A and B share at least a propositional variable or (2)
both ¬A and B are theorems.
Proof. Suppose that A → B is a theorem of RM3, but A and B
do not share propositional variables and either 0RM3 ¬A or 0RM3 B. By
completeness w.r.t. MRM3-validity (cf. [2]), 2MRM3 ¬A or 2MRM3 B.
130
GEMMA ROBLES
Then, there are MRM3-interpretations I and I 0 such that I(¬A) = 0 or
I 0 (B) = 0, that is, I(A) = 2 or I 0 (B) = 0. Suppose (1) I(A) = 2. Let
I 00 be exactly as I except that for each propositional variable pi in B,
I 00 (pi ) = I 0 (pi ). Clearly, I 00 (B) = 0 and I 00 (A) = 2 since A and B do no
share propositional variables. Thus, I 00 (A → B) = 0, whence 0RM3 A → B
by soundness w.r.t. MRM3-validity (cf. [2]), contradicting the hypothesis.
Case (2) (I 0 (B) = 0) is treated similarly. Consequently, we can conclude
that RM3 has the quasi-relevance property.
A.5. There are weak-consistent theories that are non-trivial
Let S be any extension of BRM3 included in or equivalent to RM3. We
prove:
Proposition A.6 (w-inconsistent S-theories that are non-trivial). There
are regular, prime, w-inconsistent S-theories that are, nevertheless, nontrivial.
Proof. Let pi , pm be propositional variables and consider the set y =
{B | ∃A[`S A & `S [A∧¬(pi → pi )] → B]}. It is easy to prove that y is an
S-theory. Moreover, it is regular; but y is w-inconsistent: ¬(pi → pi ) ∈ y.
So, y is inconsistent in the standard sense. Anyway, y is not trivial: for
any theorem A of RM3, [A ∧ ¬(pi → pi )] → pm is falsified by any MRM3interpretation I such that I(pi ) = 1 and I(pm ) = 0. (Recall that if A is
a theorem of RM3, I(A) = 1 or I(A) = 2 for any MRM3-interpretation
I. Cf. Definitions 2.3, 2.4.) Consequently, for any theorem A of RM3,
[A ∧ ¬(pi → pi )] → pm is not provable in RM3 (by soundness of RM3 w.r.t.
MRM3, cf. [2]). Therefore, for any theorem A in S, [A ∧ ¬(pi → pi )] → pm
is not provable in S. Thus, pm ∈
/ S, and then S is not trivial. Finally, y
is extended to the required regular, non-trivial and prime theory, by using
Lemma 3.11.
Acknowledgments
Work supported by research project FFI2014-53919-P financed by the Spanish Ministry of Economy and Competitiveness. - I sincerely thank an
anonymous referee of the RML for his (her) comments and suggestions
on a previous version of this paper.
THE QUASI-RELEVANT 3-VALUED LOGIC RM3
131
References
.
[1] A. R. Anderson and N. D. Belnap Jr., Entailment. The Logic of Relevance and
Necessity, vol I. Princeton University Press, 1975.
[2] R. T. Brady, Completeness Proofs for the Systems RM3 and BN4, Logique et Analyse
25 (1982), 9-32.
[3] R. T. Brady, Depth relevance of some paraconsistent logics, Studia Logica 43 (1984),
63–73.
[4] R. T. Brady, Hierarchical semantics for relevant logics, Journal of Philosophical
Logic 25 (1992), 357–374.
[5] R. T. Brady (ed.), Relevant Logics and Their Rivals, Vol. II, Ashgate, Aldershot,
2003
[6] R. T. Brady, Universal Logic, CSLI, Stanford, CA, 2006.
[7] J. M. Dunn, Algebraic completeness results for R-Mingle and its extensions , Journal
of Symbolic Logic 35 (1970), 1–13.
[8] J. M. Dunn, Partiality and its Dual , Studia Logica bf 65 (2000), 5–40.
[9] G. Robles, A Routley-Meyer semantics for Gödel 3-valued logic and its paraconsistent counterpart, Logica Universalis 7 (2013), 507–532.
[10] G. Robles and J. M. Méndez, The basic constructive logic for a weak sense of
consistency, Journal of Logic Language and Information 17 (2008), 89–107.
[11] G. Robles and J. M. Méndez, A Routley-Meyer semantics for truth-preserving and
well-determined Lukasiewicz 3-valued logics, Logic Journal of the IGPL 22 (2014),
1–23.
[12] G. Robles, F. Salto, and J. M. Méndez, A weak logic with the axiom mingle lacking
the variable-sharing property, Bulletin of the Section of Logic 40:3-4 (2011), 195–
202.
[13] R. Routley, R. K. Meyer, “The semantics of Entailment III”, Journal of Philosophical
Logic 1 (1972), 192–208.
[14] R. Routley, R. K. Meyer, V. Plumwood, and R. T. Brady,Relevant Logics and their
Rivals, vol. 1, Ridgeview Publishing Co., Atascadero, CA 1982.
Dpto. de Psicologı́a, Sociologı́a y Filosofı́a
Universidad de León
Campus de Vegazana, s/n
24071, León
Spain
[email protected]
http://grobv.unileon.es